Physical disk block size on POSIX using C/C++

c++, filesystems, linux, posix, unix

Solution

I wrote an answer, that while hopeful did not work correctly for block devices.

There is no POSIX mechanism for obtaining the fundamental physical block size of a device, you will have to resort to `ioctl`, which is platform dependent.

For linux there's `ioctl(fd, BLKPBSZGET, &block_size)`

For Solaris there's the `dkio` interface, which allows you to get the physical block size.

dk_minfo_ext media_info;
if (-1 != ioctl(fd, DKIOMEDIAINFOEXT, &media_info))
    block_size = media_info.dki_pbsize;

For Mac OS X, it's `ioctl(fd, DKIOCGETPHYSICALBLOCKSIZE, &block_size)`.

For FreeBSD, it should be `iotcl(fd, DIOCGSECTORSIZE, &block_size)`.

Problem

I'm working on a high performance I/O program and I'm trying to find the best way to determine the `_physical_` (and not the `_logical_`) byte size of a device's disk blocks with C++. My research so far has led me to the following code snippet: ``` #include <iostream> #include <sys/stat.h> #include <stdio.h> #include <errno.h> int main(int argc, char ** argv) { // file information including block size of the device struct stat info; // device to get block size from char * device = "/mnt/hdb1"; if (stat(device, &info)) { printf("stat() error"); strerror(errno); exit(1); } printf("Prefered block size for '%s' is %i byte\n", device, info.st_blksize); return 0; } ``` The man pages says the following about `st_blksize`: The st_blksize field gives the "preferred" blocksize for efficient file system I/O. (Writing to a file in smaller chunks may cause an inefficient read-modify-rewrite.) , but it does not mention if `st_blksize` is the logical or the physical disk block size. So, is `st_blksize` the physical disk block size, and if so, then is this the most POSIX OS portable way of detecting the physical disk block size.

Original source